PROPOSAL (Adapted from the applicant's abstract): Intractable childhood epilepsy remains difficult to treat and results in significant morbidity and cost to the individual, the family and society. With the KD, more than 50% of patients achieve clinically significant response. Despite the recent development of new antiepileptic drugs, the KD continues to play a prominent role in the treatment of intractable childhood epilepsy. The implementation and use of the KD is cumbersome and little is known about the effect of this highly restrictive diet on nutritional status. Moreover, there is limited information regarding early clinical predictors of KD response. This project will explore whether the KD implementation can be simplified without loss of efficacy, examine its impact on several key nutritional parameters, and investigate the usefulness of the electroencephalogram (EEG) as a predictor of the subjects response to the KD. In Phase 1, the current practices for initiation of the KD and its efficacy (defined as more than 50% reduction of seizure frequency at the three-month interval compared to baseline) will be investigated. The investigators propose that a gradual introduction of the KD, without a 24- to 48-hour preceding fast (Grad-KD), is as safe and efficacious as the currently recommended initiation of the KD (CR-KD) with fast. Safety of the two protocols will be evaluated during a six-day inpatient admission and the first three months of exposure. Efficacy will be determined at three months. Baseline and one-month changes in the routine EEG will be examined simultaneously. The predictive power of the changes between the baseline and one-month EEG for the three-month response (defined above) will be shown in those EEG records that have diminished background slowing or reduction of potentially epileptogenic discharges. In a pilot study, they will investigate the mechanistic hypothesis that the antiepileptic properties of the KD is due to an increase in CNS-gamma amino butyric acid (GABA), as measured by magnetic resonance spectroscopy (MRS) at baseline, 0.5-month and three-month in the CR- KD therapy. They will also examine the association between changes in CNS- GABA at 0.5 month and response to KD therapy (defined above). In Phase 2, subjects who have demonstrated a positive response at three months will be followed. The maintenance KD in responders on two caloric intakes will be evaluated. The currently recommended KD is hypocaloric, providing only 75% of subject's caloric needs, and does not allow any weight gain for the 12 months duration of treatment. The necessity of semi-starvation to maintain a positive clinical response has not been evaluated scientifically. They will examine if a eucaloric KD (Eu-KD), with adequate calories for the subject allowing for normal weight gain, is as effective as the currently recommended hypocaloric KD (Hypo-KD) with no weight gain or loss over 12 months. As part of this evaluation, comparisons of two methods used to estimate caloric needs, recommended daily allowance (RDA) tables and resting energy expenditure (by indirect calorimetry), will be made. In addition, the effect of the KD on several nutritional parameters including anthropometry, body composition and bone mineral status, will be evaluated, allowing the 12-month impact of the Eu-KD versus Hypo-KD diets on nutritional status to be compared.